Does speed actually matter in spatial biology analyses and how has the field evolved?
von Christoph Enz
Can you walk to Vladivostok?
Yes, of course you can! And it does not have to be this place, but we liked the alliteration. Just as a proverbial example of a place far away from western Europe where this is written. The journey would be roughly 10,000 km as the crow flies and probably closer to 11,500 km if you followed actual roads and walkable paths. Assuming a comfortable walking pace of 5 km/h for six hours a day, and ignoring rest days entirely, you would still spend more than a year getting there. And probably some pairs of shoes and socks worn out on the way. But it is quite possible indeed.
Modern air travel covers the same distance in less than a day. Somewhere in between sits the Trans-Siberian Railway, which would still take weeks (and yes, of course, political considerations aside).
And this is not entirely unlike the evolution of spatial biology imaging. Because in principle, you can always get your data eventually. The real question is how much time, effort, manual labour, optimisation, and suffering it takes to get there. Let us imagine a reasonably sized study cohort of 50 patients. Each patient contributes three tissue sections from a biopsy of interest, leaving us with 150 slides in total.
Now assume we want to analyze 16 biomarkers. Of course, more would always be nice. Thirty or more markers might provide even deeper biological insight, but budgets are finite, assay complexity grows rapidly, and at some point, a compromise has to be made. So let us assume 16 markers are sufficient to answer the biological question reliably.
Is imaging time actually the limiting factor?
Let us do a quick thought experiment. Our example project consists of:
- 50 patients
- 3 tissue sections each
- 16 markers
That means 150 slides which need to be imaged 16 times. Now consider that even a relatively small tissue area of 1 cm² may require roughly 500 individual fields of view to cover completely. Full slides can easily be far larger than that. Suddenly, those 150 slides turn into:
- 150 slides
- × 16 markers
- × 500 image positions
That equals roughly 1.2 million individual images. Even if each image required only 100 milliseconds of fluorescence exposure time, we would still end up with more than 30 hours of pure acquisition time.
At first glance, that may not sound catastrophic. A few days of imaging, it is not that bad, right? Well, it is bad, or rather, it would be. Because this "level 1" scenario is purely hypothetical. Simply because it cannot be done with single images on a simple FL-microscope, because not only would you have to consider the handling and manual imaging which would drive you mad, it’d also take you centuries to align the images, overlay them, and finally quantify the results, whichever fancy AI you might be able to employ. And before all that, the cycling would add hours every time, so the "getting the images" part would stretch out over months. And with so many cycles it is almost guaranteed that for the last markers there’d be fairly little tissue left to even look at.
Returning briefly to our analogy. This would be the barefoot one-legged walk to Vladivostok and close to impossible. You would have to prioritize the journey over the destination quite a lot to enjoy it.
What have we learnt so far? The imaging time as such is probably not the limiting factor. So, let’s move on to a more realistic scenario. The minimum equipment - proper hiking shoes, so to say - would have to be a digital fluorescence microscope with a standard 4-channel setup and automated stage positioning and storage. If your lab is well equipped it could be reasonably fast, have a really good (maybe confocal) image quality and come with a decent data analysis software (By the way, if you do not have anything like this yet but are on the lookout, please have a closer look at our website. Not so much for the spatial work, where we can recommend something special down below, but for general high quality high content imaging).
The workflow would still remain cyclic. You would stain the tissue with antibodies for the first set of markers, image the section, remove or bleach the fluorophores, and repeat the process with the next marker set. And again of course, hope that your tissue does not suffer on the way (see our blog on "why cyclic methods can make you suffer").
Imaging itself may only take about a minute per slide and cycle. Even with four cycles required for 16 markers, pure acquisition time can remain surprisingly manageable. The real limitation lies in the surrounding workflow. Cycling steps often limit throughput to perhaps one complete processing round per day. Even if dozens of your 150 slides are imaged within a morning, much of the remaining day may disappear into staining, washing, bleaching, handling, and quality control. But hey, you’ve made progress, and with a little luck and within several weeks you may have collected the images.
Unfortunately, image acquisition is still only part of the story. All generated image sets must still be aligned, overlaid, segmented, analyzed, and quantified. Depending on software quality, computational infrastructure, and user experience, this processing phase alone may take months rather than just weeks, which may let you finish the above-mentioned project in the best case inside a half year timeframe. And before any of this even starts, assay development itself requires substantial effort. Antibodies must be selected, tested, optimized, validated, and combined into functioning panels. Each individual marker may require days or weeks of optimization work, and multiplex compatibility adds another level of complexity. By the time the actual study begins, considerable amounts of time and money may already have been invested.
Still, compared to our original scenario, we have moved from “almost impossible” to “difficult but manageable.” This would be your walk to Vladivostok.
However, since “technically feasible but painfully labour-intensive” was not an ideal long-term solution, the industry spent much of the past decade developing increasingly integrated spatial biology platforms. And hence has taken us to level 3.
These systems improved automation considerably. Pre-validated antibodies, integrated workflows, dedicated software environments, and improved analysis pipelines reduced both preparation time and user workload. Data evaluation also became faster and more reproducible. What largely remained unchanged, however, was the reliance on cyclic workflows. As soon as larger marker panels are involved, repeated staining and imaging rounds still dominate the process. Some systems additionally rely on secondary antibody strategies, which can further increase cycle numbers, processing times, and sometimes even compromise data quality (please see our blog on "caveats in using secondary antibodies in spatial biology imaging").
As a result, imaging times improved somewhat, but complete projects of our example size would still typically require several weeks for acquisition alone. And every additional cycle introduces another layer of image overlay, sucking up computer power, calculation time, and potentially precision in the result. And also, part of the truth, the actual preparation and assay development times continue to vary substantially between system providers and depend heavily on the size of validated antibody catalogues, the ease of introducing additional markers, system robustness, and the quality of vendor support.
Compared to earlier generations, these systems represented major progress. But the field was still largely walking, just with much better hiking boots and the occasional stage on the Trans Siberian. Fortunately, the field did not stop there. We would not have dragged you through levels 1-3, if there wasn’t a level 4 now. And we can’t help getting a little "marketing‑ish" here, as the progress is quite spectacular.
With Rarecyte’s introduction of the Orion HT system some more "researchy" types of projects are now down to pure routine. Which is, what projects of our model size should be. You will find more on the Orion HT on our respective website, since we do not want to bore you with specs here.
Only that instead of relying heavily on repeated cyclic processing, up to 20 markers can now be acquired within a single imaging round. Automation enables unattended processing of up to 30 slides per run, while our 1 cm² tissue section with 16 markers can be imaged in little more than an hour.
For our example project of 150 slides, this means that the entire imaging phase can realistically be completed within a week. At the same time, an extensive catalogue of pre-labelled and validated antibodies dramatically reduces assay development effort. Dedicated panel-building software simplifies experimental setup further, while proprietary and simple labelling chemistry still allows users to introduce additional custom markers when necessary.
The one-shot approach also eliminates much of the fluidics complexity associated with cyclic workflows, which everyone who has ever dealt with microfluidic peculiarities might appreciate for its implications for the general robustness of just any system. And finally, integrated analysis environments ensure that the output is not merely visually impressive imagery, but robust quantitative biological data.
So, does speed matter?
Yes, but perhaps not in the way many people initially assume. Pure image acquisition time is often not the main limiting factor. Workflow complexity is. The real challenge lies in minimizing manual intervention, reducing cyclic processing, simplifying assay development, maintaining tissue integrity, and making large-scale analysis reproducible and practical.
The goal is to reach actionable biological conclusions within meaningful timelines.
So, if not the journey but the destination is what you´re after, you may prefer flying to Vladivostok over walking for months on end.